In addition to designing and training machine learning models, we will also stay around afterwards to make sure the model is running as it should and can be effectively integrated into your business infrastructure.
It often takes as long to deploy a machine learning model as it took to build it, especially in a large company.
Considerations to take into account in deployment are:
- How should the model communicate? Over an API, web request, etc?
- Do we want a cloud hosted or on premise solution? If cloud hosted we have a choice between standard server or serverless compute, which come with their own pricing structures.
- Speed, load, how many instances, is a load balancer needed?